One identity, one source of truth, one approval graph.

Hartzell has built, painted, and protected South Florida properties since 1948. Over more than seven decades it grew into three businesses, construction, painting, and impact windows, run by one team. That scale is exactly what its software could no longer keep up with.

The before state

Before us, the work was scattered across roughly a dozen disconnected systems, and no two of them agreed: a decades-old CRM that was barely holding together, separate accounting and ERP platforms, a standalone document-storage system, a construction project-management tool for the field, and a dedicated HR and payroll system. Around all of that was a layer of shadow IT: spreadsheets on a dozen desktops, project boards a few teams kept on their own, calendars scheduled by hand, an outside marketing vendor holding its own data, and, genuinely, a stack of printed pages and notes written on napkins.

Every customer-facing motion meant a person copying data from one system into another. Onboarding a new hire touched several of them and took about two days of admin. There was no single source of truth and no audit trail.

The pitch most firms would have led with, "let's add an AI agent," doesn't work here. When the systems can't talk to each other, the agent can't see the data, take the action, or write the audit. Fragmentation wins.

What we did

We didn't bolt the old tools together. We replaced them with one platform we built, on a single identity, a single source of truth, and a single approval graph.

Unify first. One identity plane (hiring someone provisions their access everywhere; offboarding is one click), one system of record, one approval graph for every sign-off, and one encrypted secret store. Every record reconciled into one place.

Then build what the business actually runs on. On that foundation we built the operations software itself: estimating, field inspections, project management, and accounting, all sharing the same data instead of fighting it. We used cutting-edge AI to build it, with methods of our own that go well beyond prompting a chatbot, which is how a platform this size reached production in a fraction of the usual time without giving up rigor: the code is tested, owned by the client, and theirs to keep.

Then add AI inside it, only where judgment lives. A document bot reads every incoming invoice and bill, pulls the line items, and assigns each one to the right project automatically, with anything uncertain routed to a person. That invoice-reading layer is the same thing we build for other contractors as custom construction invoicing software. AI pulls fields from uploaded contracts and insurance documents. It never touches the approval path itself: approvals stay deterministic, role-attributed, and auditable.

We use cutting-edge AI to build these systems, and we are deliberate about where AI runs inside them. Unify first, then automate, then add AI only where judgment lives. Run it in any other order and it stalls.

The outcome

• Employee onboarding: about two days of admin down to minutes, driven by identity.
• Subcontractors: a shared spreadsheet and an email chain replaced by one self-serve portal.
• Invoices and bills: sorted by hand before, now read by AI and auto-assigned to the right project, with exceptions sent to a person.
• Every approval: timestamped, role-attributed, and queryable.

Because every part of the platform shares one identity, one source of truth, one approval graph, and one secret store, a new module drops onto the same substrate in days, with AI added where it earns its place, not bolted on top.